A prototype earns its place when it helps you learn something that could change what you build.
Name the risk first.
Before making screens, write the assumption that would make the idea fail. It might be that customers do not understand the value, cannot complete a crucial task, do not trust an automated answer, or will not change their existing workflow.
Choose one assumption. A prototype that tries to answer every question often becomes an expensive demonstration.
Match fidelity to the question.
- Value: a concept description or storyboard may be enough to discuss the problem and alternative approaches.
- Usability: a clickable core journey can reveal confusing decisions or missing information.
- Feasibility: a technical spike may be more useful than polished screens.
- Trust: realistic examples, explanations, and error states can expose what people need before acting.
Include the difficult moment.
AI-assisted prototyping makes a happy path easy to produce. Deliberately include an empty state, an uncertain output, an edit, or an error if that is where the real risk lives. Show how the person checks, corrects, or rejects a result.
Write the task without giving away the answer.
“Click the invite button” tests little. “You need your colleague to review this project, but they should not be able to change it” reveals whether the experience supports the person's goal.
Record what success would look like before the session. Watch behavior and ask neutral follow-up questions. Avoid turning every hesitation into a request for a new feature.
Decide what the result means.
Before testing, write three possible decisions: continue with the current approach, change a specific part, or stop and revisit the problem. Identify what evidence would lead you to each.
A small qualitative study can reveal issues and improve your understanding. It does not establish a reliable population conversion rate. Use later behavioral experiments where that question matters.
Keep the handoff honest.
Label simulated data and interactions. Document what is real, what is mocked, what remains technically uncertain, and what must be tested in production. A convincing prototype should not quietly become a promise that all of its behavior is ready to ship.
Put this into practice.
Start with the free worksheet or talk with Ashley about your specific challenge.